 | |||
|
|
||
| |||
| |||
| |||
|
||||||||||||||||||||
| |
|
Home > Data management / BI News > Teradata VP talks data warehouse appliances, reveals cloud and SSD plans |
| |
|
|
|
|
| Data management / BI News: |
|
||
|
|||
|
|||
|
|||
|
|||
|
|||
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||
With the data warehouse appliance market heating up, traditional data warehousing heavyweight Teradata threw its hat into the appliance ring last year with three new products. But to hear Teradata tell it, the vendor has been in the enterprise data warehouse appliance game for more than 25 years. "If you define an appliance as a built-for-a-purpose platform, totally integrated – disks, CPU, storage, database, etc. -- from a single vendor, we've actually been an appliance vendor since 1984," said Randy Lea, Teradata Corp.'s products and services marketing chief. "So in many regards we were the original appliance vendor," Lea said. Still, in an interview with SearchDataManagement.com, Lea acknowledges Teradata had, until last Spring, largely ignored the market for small – read: non-enterprise-level – data warehouse deployments. "We didn't have an affordable product if you were just doing workgroup-type data warehousing," he said. Lea also discussed why Teradata continues to use proprietary MPP hardware when many of its competitors have turned to commodity servers. And he shares his views on data warehousing in the cloud; Teradata's competitors Oracle, IBM and soon to be competitor Microsoft; and on the company's plans for solid-state drives. Following is an edited version of Lea's interview with SearchDataManagement.com. SearchDataManagement.com: Teradata got into the data warehouse appliance game a little over a year ago with the introduction of theTeradata 550, 2550 and 5550 appliances. Why did Teradata decide to do this and is the company shifting its focus to the appliance model? Randy Lea: What we wanted to do is, number one, extend our market share in our existing accounts. Though the majority of our accounts are building enterprise data warehouses, they also build data marts, and we never really went after that business. So we thought, well, we have the technology, we have the database.
That said, we're not losing our focus on the Active Enterprise Data Warehouse platform. In fact, that's still our primary enterprise product. And we're going to continue to enhance it to increase its functionality as well. Data warehouse appliances are just a way for us to expand our market reach. SearchDataManagement.com: Data warehousing in the cloud – either public or private – is getting a lot of attention these days. What is Teradata's position on cloud computing for analytic databases? Lea: We think that the cloud has tremendous potential. We're actually working with Amazon today and doing some testing on the technology. We're also testing in VMware environments to see how Teradata reacts in a shared, virtualized environment. We also have our own internal cloud offering that we'll be releasing at our Partners Conference. That's the ability to set up agile analytic clouds within your enterprise data warehouse, so if you have a new user that wants to experiment and play around with some data, we can actually provision space for them within the Teradata data warehouse environment. You can play around with it and if you want to productionize it, you can. Ebay does that very effectively as well as some of our other customers. Quite frankly, we'll now take advantage of the cloud hype, if you will, and let customers do this internally on Teradata. SearchDataManagement.com: So it sounds like Teradata is quite bullish on the cloud computing phenomenon? Lea: Well, when you go to the cloud, public or private, there are some technical issues with all data warehouse environments whether it's Teradata or anything else. One of the challenges of why you separate an OLTP environment from a data warehouse environment is that OLTP handles different workloads typically on a single application with no queries, no performance requirements, etc. A data warehouse is dynamic, and in a parallel environment, which is different than most OLTPs, you have to run all of your executions in parallel. And for the most part, with every parallel technology, if one of your parallel units runs slower than the others, you pretty much run at the slower execution. So, when you put yourself on a cloud [where] there are typically no guaranteed service levels, now all of a sudden you're spreading your parallelism across some of these units of power that have really no dedicated service levels. Now your performance is somewhat dependent on the resources you get at that time. In a public cloud, the performance availability can be somewhat dramatic. If I'm just doing some deep analytics and I just want a cheap resource to learn something that I don't care if I get the answer back for a day, it'll work just fine. But if I have to run my business with guaranteed service levels, especially in the active environment, the cloud is probably not something I can use at this point in time. SearchDataManagement.com: In Gartner [Research]'s latest Magic Quadrant for Data Warehouses, Teradata was at the top of the leaders' quadrant. Not far behind are three mega-vendors – Oracle, IBM and Microsoft. What do you make of your competition, and do the mega-vendors have an advantage with complete data management stacks to offer customers versus Teradata, which focuses solely on data warehousing? Lea: Some analysts have predicted they will catch us in our leadership position in two years, but they've been saying that for 15 years. I think the main reason for that is our focus. Every R&D dollar we put in is around data warehousing. We respect our competition, but in the database space, we feel very confident. In the BI tools space and the data integration space, they're some of our best partners. Many of the acquisitions that the mega-vendors have made were actually strong partners of ours prior to the acquisition: Siebel Analytics by Oracle; Cognos by IBM. We still have those strong relationships. We still have engineering-to-engineering partnerships. They're in our R&D facility probably about twice a year. And if you think about it, Oracle, IBM and SAP with Business Objects paid good money for those products, and if they isolated them just to work on their database, there's no way they could justify the cost of the acquisition. They have to be somewhat database agnostics. So, we meet with them and they actually build in features that are optimized for Teradata. In fact, most of them have settings that if you're running on Teradata, the software will act this way versus other databases. SearchDataManagement.com: What about competition from the smaller pure-play vendors? Often, it is the challengers like these that drive innovation. Do you think that's the case in the data warehousing space? Lea: I think the smaller pure-play vendors bring in an interesting element in that they focus more on the workgroup. As I said, that was an area that we weren't really focused on. So, we kind of thank them for giving us the opportunity to look at a market that quite frankly we didn't necessarily look at as being attractive enough for us. Long term, though, it's a tough market to establish yourself, especially with Oracle Exadata putting pressure on some of those players. And our appliance family is definitely putting pressure on those players. We don't see any technology they're bringing to the table that is unique or differentiated or earth shattering. SearchDataManagement.com: So what can customers expect from Teradata in the months and years ahead in terms of innovation? Lea: I think you'll see some interesting products and announcements around solid state drives in the future. We demonstrated the first prototype data warehouse platform with solid state drives at Partners last year, and we see that technology continuing to improve and the price point coming down. So when it makes sense where solid state drives come down in price and there's a nice price performance curve, we're ready. We also recently announced Teradata Virtual Storage. This is automatic data storage based on temperature. What we mean by temperature is access rates. If data is accessed a lot, then it's 'hot' and if it's not accessed that much, it's 'cold.' We will actually move data to hotter placements on existing drives. So, now, we can put data on the outer portion of the drive, which is faster than the inner part of the spindle. In the future when we have solid state drives or maybe large static drives, we will actually move data between drives for different performance levels.
|
|
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||
| |||||||||||||||||||||||||||||||||||||||||||||||||||||||
|
|
About Us | Contact Us | For Advertisers | For Business Partners | Site Index | RSS |
|
|
||
|
|||||||
